Designing Representative and Balanced Experiments

Randomized controlled trials (RCT's) are increasingly common in applied microeconomics research. For a fixed budget, researchers can maximize precision by carefully randomizing in a way that finely balances unit-level covariates. Stratified rerandomization is one such covariate-balancing design. It works by (1) matching units into homogeneous groups then (2) iteratively randomizing within these groups until balance is achieved. This project will work on a stable and well-documented Python package implementing this design, including post-randomization covariate adjustment and inference methods. Time permitting, the student will also run simulation experiments for new covariate balancing designs.